Consciousness and Cognition 22 (2013) 875–886

Contents lists available at SciVerse ScienceDirect

Consciousness and Cognition

journal homepage: www.elsevier .com/locate /concog

Conscious recollection and binding among context features

1053-8100/$ - see front matter � 2013 Elsevier Inc. All rights reserved.http://dx.doi.org/10.1016/j.concog.2013.05.010

⇑ Corresponding author. Address: Department of Psychology, University of Mannheim, D-68131 Mannheim, Germany. Fax: +49 621 181 2129E-mail address: boywitt@uni-mannheim.de (C.D. Boywitt).

C. Dennis Boywitt ⇑, Thorsten MeiserSchool of Social Sciences, Department of Psychology, University of Mannheim, Germany

a r t i c l e i n f o

Article history:Received 19 February 2013Available online 19 June 2013

Keywords:Conscious recollectionRemember/knowEpisodic memorySource memoryContext binding

a b s t r a c t

Recent research suggests that the subjective feeling of conscious recollection is uniquelycharacterized by joint memory for several context features while merely familiar memorieslack this property (Meiser, Sattler, & Weisser, 2008). In the present research we took thenovel approach of extending the dual task paradigm to the simultaneous study of subjec-tive retrieval experience (using the remember/know procedure) and joint memory for twoorthogonal context features. While dual task load during encoding lead to reductions in thefrequency of the subjective experience of conscious recollection and reductions in overallcontext memory, joint context memory was not affected. Furthermore, the relation ofhigher overall context memory for consciously recollected items than for familiar itemswas preserved even under dual task load. These results have import implications for the-ories of long-term feature binding and the processes involved in producing the experienceof conscious recollection.

� 2013 Elsevier Inc. All rights reserved.

1. Introduction

Episodic memories are defined by the retrieval of contextual details specifying the circumstances under which theseevents were encoded and the associated subjective experience has been termed conscious recollection. Memories whichare accompanied by a mere feeling of familiarity, on the other hand, are not thought to be associated with the retrievalof episodic details (Tulving, 1985). While there is empirical evidence that conscious recollection is indeed associated withbetter memory for contextual details than is a feeling of familiarity (Dewhurst & Hitch, 1999; Dudukovic & Knowlton,2006; Meiser & Bröder, 2002; Perfect, Mayes, Downes, & VanEijk, 1996), recent research has demonstrated that consciousrecollection is more specifically characterized by joint memory for different context details from the encoding episode (Boy-witt & Meiser, 2012a, 2012b; Meiser & Bröder, 2002; Meiser et al., 2008).

These recent studies show a consistent relationship between the subjective retrieval experience (as assessed with theremember/know paradigm; Tulving, 1985) and joint retrieval of the context details such that ‘remember’ responses wereassociated with joint memory of the context features whereas ‘know’ responses were not. The process underlying jointmemory of different context features was termed context–context binding (Meiser & Bröder, 2002). As opposed to simpleitem-context binding, context–context binding may be conceived as a stronger concatenation of the context features inthe memory trace, increasing the probability of joint recall of context features (Boywitt & Meiser, 2012a). In line with theconceptualization of episodic memories as integrated representations of the study episode (e.g., Parkin & Walter, 1992;Perfect et al., 1996; Rajaram, 1993; Tulving, 1985) these studies suggest that memories which are experienced as consciouslyrecollected are coherent re-instantiations of the feature configuration at study while memories which feel only familiar lackthis property. The present research pursued two interrelated goals. First, we aimed to fill the gap in knowledge of the

.

876 C.D. Boywitt, T. Meiser / Consciousness and Cognition 22 (2013) 875–886

underlying processes involved in creating context–context binding as a correlate of recollective experience. Second, wesought to shed light on the relation of conscious recollection and context memory in general.

2. Mechanisms of context–context binding

Previous research suggests that context–context binding is produced by an encoding process operating in the immediatefocus of attention which integrates the features of the episode into a coherent memory representation (Boywitt & Meiser,2012b; Uncapher, Otten, & Rugg, 2006). Evidence for this view comes from an imaging study indicating that the intra-pari-etal sulcus, which is thought to be involved in attentionally mediated binding, is especially activated during encoding ofitems with two independent context features which are later remembered with both context features as opposed to itemswhich are later remembered with only one context feature (Uncapher et al., 2006). Uncapher et al. suggested that allocationof attention to the ‘‘object level’’ increases the likelihood of establishing bound representations. Behavioral studies offer con-verging evidence that context–context binding is only observed for context features which are in the focus of processing dur-ing encoding (Boywitt & Meiser, 2012b). In an incidental learning situation, context–context binding was only observed forstimulus-bound (i.e., intrinsic) context features (i.e., font color) but not for more peripheral (i.e., extrinsic) context features(i.e., color of a frame), which are not in the immediate focus of attentional processing (Ecker, Zimmer, & Groh-Bordin, 2007;Mulligan, 2011). When intrinsic and extrinsic context features were in the focus of processing due to explicit learninginstructions, however, context–context binding was observed for both intrinsic and extrinsic context features, suggestingthat context–context binding depends on an attentionally mediated encoding mechanism.

Further evidence comes from research on visual short-term memory, suggesting that binding of different perceptual fea-tures is mediated by a resource-demanding process (e.g., Brown & Brockmole, 2010; Rensink, 2000; Stefurak & Boynton,1986; Treisman & Gelade, 1980; Treisman & Schmidt, 1982; Wheeler & Treisman, 2002). To illustrate, in a visual short-termmemory task, Wheeler and Treisman (2002) presented participants with multiple objects simultaneously, which differed inlocation and color. While change detection performance for the independent features was quite good, changes that affectedthe binding between location and color such as re-arranged pairs of previously presented features were detected at a muchlower rate. Additional empirical evidence for a resource-demanding binding process comes from a study in which partici-pants were required to perform a mental arithmetic task during the retention interval between the presentation of coloredshapes and the test phase (Stefurak & Boynton, 1986). This manipulation affected the detection of new shape-color pairingswhile accuracy for the individual features was largely preserved. In a similar vein, Treisman and Schmidt (1982) have shownthat the rate of conjunction errors (i.e., false recognition of re-arranged features from different objects) increased under cog-nitive load during presentation of the stimuli. Similarly, another study found verbal–spatial binding to be impaired by a con-current working memory load during stimulus presentation (Elsley & Parmentier, 2009), indicating that the integration ofmultiple features relies on a resource-demanding process.

Although there is ample evidence that feature binding may rely on resource-demanding processes, there is also evidencesuggesting that binding of different features might not be resource-demanding over and above the demands of encoding andmaintaining the individual features (e.g., Allen, Baddeley, & Hitch, 2006; Delvenne, Cleeremans, & Laloyaux, 2010; Luck &Vogel, 1997). Allen et al. (2006), for example, did not find decreased recognition performance in a short-term memory taskfor combinations of features under additional load as compared with the features alone.

3. Goals of the present research

Because previous research has provided only indirect evidence on the attentional demands of context–context binding,we tested the prediction that binding of context features depends on resource-demanding processes (Boywitt & Meiser,2012a, 2012b; Uncapher et al., 2006) directly by implementing a dual-task manipulation. For that purpose, we employeda multidimensional source memory paradigm with two completely crossed context dimensions (Meiser & Bröder, 2002)and assessed subjective retrieval experience with the remember/know-procedure. In the multidimensional source memoryparadigm participants study items varying on two orthogonal context dimensions, for example items presented at two dif-ferent screen positions and two different font sizes. Assessing memory for two completely crossed context dimensions al-lows for the analysis of stochastic dependence in context memory. For instance, if retrieving one context feature (e.g.,font size) is more likely if the other context feature (e.g., screen position) is also retrieved compared with retrieving the fea-ture font size alone, then context memory is stochastically dependent. The phenomenon of stochastic dependence in mem-ory for context features has been taken as an indication of context–context binding.

Theoretically, stochastic dependency in memory for two orthogonal context features is independent from ‘‘simple’’ mem-ory for each context feature. That is, memory could be fairly accurate for both context features but may still be stochasticallyindependent (cf. Light & Berger, 1976). Indeed, this pattern of results has been observed when ‘‘simple’’ context memory wasexperimentally equated between R and K judgments but stochastic dependency was still only observed for R responses, sug-gesting a qualitative difference in the underlying memory characteristics of R and K responses (Meiser et al., 2008).

In the first experiment stochastic dependency in context memory as an indication of context binding was compared be-tween a condition with dual task (DT) load during study and a full attention condition. In the second experiment we broad-ened our focus to include the relation of subjective retrieval experience and context memory in general.

C.D. Boywitt, T. Meiser / Consciousness and Cognition 22 (2013) 875–886 877

4. Experiment 1

If, as much of the literature on feature binding suggests, context–context binding also relies on a resource-demandingencoding process then DT load during encoding should diminish context–context binding. Alternatively, context–contextbinding might be spared by DT load during encoding, suggesting that the underlying process is rather automatic and/or oc-curs during retrieval.

Complicating the matter, previous research suggests that DT load reduces overall context memory (e.g., Craik, Luo, & Sak-uta, 2010; Troyer, Winocur, Craik, & Moscovitch, 1999) and thus, a reduction in context–context binding under DT loadwould be confounded with lower overall context memory. In order to avoid such a confound we equated overall contextmemory between the DT group and the control group by making the context features more difficult to distinguish for thecontrol group. More similar context features have previously been shown to impair context memory (e.g., Bayen, Murnane,& Erdfelder, 1996), so that the manipulation is suitable to equate overall context memory across groups.

4.1. Method

4.1.1. ParticipantsNinety-five students (mean age = 22.6, SD = 4.29 years; 71.6% female) participated in this study for either course credit or

monetary compensation and were randomly assigned to the two groups. Forty-nine participants were assigned to the controlgroup and 46 participants were assigned to the DT group.

4.1.2. Materials and designFrom a set of 165 concrete German nouns of four to seven letters, 64 words were randomly drawn as study items and 40

words were randomly drawn as distracters for the test phase for each participant anew.Because the dimensions font size (large vs. small) and position on the screen (left vs. right) were used in several previous

studies (e.g., Boywitt, Kuhlmann, & Meiser, 2012; Meiser & Bröder, 2002; Meiser et al., 2008; Riefer, Chien, & Reimer, 2007;Starns & Hicks, 2008) we used the same context dimensions in order to be consistent with previous research and to allowcomparisons with previous results. In order to equate context memory across groups, the context features were more similarfor participants in the control group than for participants in the DT group. During the study phase, half of the 64 study itemsappeared on the left side of the screen and the other half appeared on the right side of the screen. Orthogonal to the positionon the screen, the words varied with regard to font size. For participants in the DT group, the distance between the centers ofitem presentation was 19 cm whereas for participants in the control group the distance was only 15.2 cm. For participants inthe DT group, the larger font size was 2.2 cm letter height and the smaller was 0.9 cm letter height. For participants in thecontrol group, the letter heights were 1.5 cm and 0.9 cm, respectively. To balance the appearance of both features across thestudy list, the list was divided into four blocks of 16 items. In each block, each combination of the features location and fontsize appeared equally often in an order that was randomized for each participant anew. Presentation duration was three sec-onds with an ISI of one second in both groups.

For the secondary task a tone monitoring task was employed for which participants were presented with a random se-quence of two tones differing in pitch (440 Hz and 330 Hz). Each tone lasted for one second followed by an ISI of two sec-onds. The participants were instructed to press the space bar on the keyboard whenever they heard three consecutive tonesof the same pitch. This task should tap central executive resources without interfering with the visual modality in which thestimuli were presented (Gardiner & Parkin, 1990; Parkin & Russo, 1990).

4.1.3. ProcedureInstructions, presentation of materials, and the memory test were computer directed. The initial instructions were iden-

tical for the group with the tone monitoring task and the control group and informed participants that they would be pre-sented with a series of words which were to be memorized for a later memory test. Importantly, participants were notinstructed to memorize the font size and location of the words. Participants in the group with the tone monitoring task werethen informed about the additional task and were given a short practice block of 8 tones. Following the instructions, bothgroups were presented the 64 study items one at a time. After the study phase, all participants were asked to write downas many European capitals as possible for three minutes. Subsequent to this filler task all participants received instructionsconcerning the memory test. Then the combined old/new, remember/know, context memory test commenced. During thetest phase, target items and distracters were displayed in random order in the center of the screen in a font size intermediatebetween the large font size and small font size used in the study phase. All responses to the memory test were made by click-ing with the mouse on response fields on the screen. For each item, participants were first asked to indicate whether theword was ‘‘old’’ or ‘‘new’’. If participants selected ‘‘old’’, they were asked to indicate whether they ‘‘remember’’ or ‘‘know’’the word. Then, participants were asked to indicate in which combination of font size and position on the screen theyhad studied the word. For that response, four response fields with the four combinations of font size and location were dis-played. The instructions for the remember/know judgments were identical to the instructions used in previous studies (Boy-witt & Meiser, 2012b; Meiser & Bröder, 2002) and avoided any explicit association of R responses with the recollection ofperceptual details like font size or location.

878 C.D. Boywitt, T. Meiser / Consciousness and Cognition 22 (2013) 875–886

4.2. Results

An alpha level of .05 was specified for all analyses.

4.2.1. Remember-know judgmentsTable 1 shows the proportion of R and K responses to target items (i.e., hits) and distracters (i.e., false alarms).A two by two mixed ANOVA with type of judgment (R vs. K) as within participants factor and group (DT vs. control) as

between participants factor on hits revealed a significant main effect of type of judgment, F(1,93) = 23.60, MSE = 0.05,g2

p = .20, p < .001, with higher rates of hits accompanied by R than K judgments. The main effect of group was also significant,F(1,93) = 8.21, MSE = 0.02, g2

p = .08, p = .005, indicating lower hit rates for the DT group than the control group. The interac-tion was marginally significant, F(1,93) = 3.37, MSE = 0.05, g2

p = .04, p = .07.Independent t-tests revealed that the rate of R responses was lower in the DT group than in the control group, t(93) = 2.56,

d = .52, p = .012, but rates of K responses did not differ between groups, t(93) = 0.17, d = 0.03, p = .887.The same two by two mixed ANOVA on false alarms revealed a significant main effect of type of judgment,

F(1,93) = 53.30, MSE = 0.01, g2p = .36, p < .001, with higher false alarm rates for K responses than for R responses. Neither

the main effect of group nor the interaction was significant, F(1,93) = 1.70, MSE = .01, g2p = .02, p = .20, and F(1,93) = 0.42,

MSE = 0.01, g2p = .004, p = .52, respectively.

4.2.2. Overall context memoryNext we tested whether overall context memory differed between the two groups. To measure overall context memory,

we computed for all correctly recognized target items the probability of assigning the items to their respective context foreach context dimension (i.e., font size and location) separately. Table 2 displays the average conditional source-identificationmeasure (ACSIM; Murnane & Bayen, 1996) scores. One-sample t-tests revealed that context memory was above chance level(i.e., >.5) for R responses in both groups (see Table 2).

A mixed ANOVA on ACSIM scores with the within participants factors context dimension (position vs. font size) and judg-ment (R vs. K) and the between participants factor group (control vs. DT) revealed a main effect of judgment,F(1,88) = 105.75, MSE = 0.03, g2

p = .55, p < .001, with higher context memory for R than for K responses. The main effect ofcontext dimension was marginally significant, F(1,88) = 3.90, MSE = 0.02, g2

p = .04, p = .051, with slightly higher contextmemory for position than for font size. Neither the main effect of group was significant, F(1,88)= 0.21, MSE = 0.04,g2

p = .002, p = .65 nor any of the interactions: Judgment by group, F(1,88) = 1.17, MSE = 0.03, g2p = .01, p = .282, Judgment by

context dimension, F(1,88) = 2.87, MSE = 0.02, g2p = .03, p = .094, context dimension by group, F(1,88) = 0.69, MSE = 0.02,

g2p = .01, p = .407, three-way interaction, F(1,88) = 0.02, MSE = 0.02, g2

p < 0.001, p = .904. Thus, context memory was success-fully equated between both groups.

4.2.3. Model-based analysesIn the next step we analyzed the responses in the memory test with a re-parameterized variant of the multinomial pro-

cessing tree model for two crossed source dimensions (Meiser, 2009; Meiser & Bröder, 2002). The model-based analyses offer

Table 1Mean proportions (and standard errors) of remember and know responses to target items (hits) and distracters (false alarms) in Experiments 1 and 2.

Experiment 1 Experiment 2

Control Dual task at encoding Control Dual task at encoding Dual task at retrieval

HitsRemember .47 (.03) .35 (.03) .51 (.03) .33 (.03) .52 (.03)Know .24 (.02) .25 (.02) .21 (.02) .27 (.02) .24 (.02)

False alarmsRemember .03 (.01) .04 (.01) .02 (.01) .04 (.01) .02 (.004)Know .11 (.02) .14 (.02) .05 (.01) .14 (.01) .08 (.01)

Table 2Mean conditional source identification (ACSIM) scores (and standard errors) in Experiments 1 and 2.

Experiment 1 Experiment 2

Control Dual task at encoding Control Dual task at encoding Dual task at retrieval

RememberPosition .63 (.02)a .65 (.03)a .72 (.02)a .65 (.02)a .71 (.02)a

Font .59 (.02)a .59 (.02)a .64 (.02)a . 64 (.02)a .60 (.02)a

KnowPosition .45 (.02) .43 (.03) .63 (.03)a .50 (.02) .57 (.02)a

Font .46 (.03) .42 (.02) .54 (.02) .56 (.02)a .55 (.02)a

a Value significantly above .5 (p < .05).

C.D. Boywitt, T. Meiser / Consciousness and Cognition 22 (2013) 875–886 879

a variety of advantages over traditional analyses. First, these analyses allow for the simultaneous estimation of latent param-eters for recognition memory, the subjective feelings of conscious recollection and familiarity, source memory, responsebiases, and guessing processes. Second, the model allows for the estimation of latent parameters indicating independentmemory for each source dimension and for the analysis of stochastic dependency in source memory. The parameter esti-mates are frequently preferred over behavioral measures which may be contaminated by nuisance processes such as guess-ing and response tendencies (Batchelder & Riefer, 1990; Bayen et al., 1996; Erdfelder et al., 2009).

Fig. 1. Processing-tree diagram of the re-parameterized variant of the multinomial model of multidimensional source memory (Meiser, 2009; Meiser &Bröder, 2002), taken from ‘‘The role of source memory in older adults’ recollective experience’’ by C.D. Boywitt, B.G. Kuhlmann, and T. Meiser (2012),Psychology and Aging, 27, 484–497. Adapted with permission.

�A

mer

ican

Psyc

hol

ogic

alA

ssoc

iati

on20

11

880 C.D. Boywitt, T. Meiser / Consciousness and Cognition 22 (2013) 875–886

Consider Fig. 1 for the following explanation of the models’ parameter. The stimuli on the left side of Fig. 1 are connectedwith the observed response on the right side via the latent cognitive states represented by the model parameters. Given thatthe participant is presented with a previously studied item (i.e., ‘‘target item’’) at test (upper tree in Fig. 1), the parameter Ddenotes the probability of recognizing this target item as ‘‘old’’. The parameter R represents the probability that this recog-nized target item is judged as ‘‘remembered’’ whereas the complementary probability 1 - R represents the case when a rec-ognized target item is judged as ‘‘known’’. Thus, this parameter indicates the propensity with which recognized items areexperienced as ‘‘remembered’’ controlling for item recognition performance (parameter D) and response biases (parametersb and R�). Accordingly, all subsequent parameters are specified conditional on the ‘‘remember’’/’’know’’ distinction as indi-cated with the superscript R or K. Given that participants do not recognize a target item as old (1 � D) they may still guessthat this item is old (b). The parameter b thus denotes the probability of guessing that an unrecognized old item or an unde-tected distracter is old. In this case participants may then erroneously give a R response with probability R� and a K responsewith probability 1 � R�.

The parameters dR and dK represent the probabilities of joint memory for both source dimensions over and above the con-tribution of independent source memory (parameters eDim1R, eDim2R, and eDim1K, and eDim2K). Accordingly, if dR > 0 or dK > 0then memory for both context features is stochastically dependent indicating context–context binding. Thus, unlike theanalyses of the behavioral source memory measure (ACSIM), the model-based analyses allow us to specifically measure con-text binding for R and K responses in terms of stochastic dependence in context memory. If dR is smaller in the DT group thanin the control group, this suggests that context–context binding was reduced by DT load over and above any effects on R andD. If dR does not differ between the DT and the control group, on the other hand, context–context binding was not dissociatedfrom R responses.

Last, the guessing processes in source memory decisions when source-specifying features cannot be remembered are cap-tured by the parameters aDim1R

jA=B , aDim2RjA=B , aDim1K

jA=B , aDim2KjA=B given a studied item was classified as R or K, respectively. The parameters

gDim1RjA=B , gDim2R

jA=B , gDim1KjA=B , and gDim2K

jA=B capture the guessing processes in source attributions if a study item was not detected as old(1 � DN) but was still guessed to be old (with probability b).

Given the participant is presented with a new item at test (i.e., ‘‘distractor item’’ in the lower tree in Fig. 1) the parameterDN denotes the probability of identifying this distracter item as ‘‘new’’. Given that participants do not detect a distracter item(with probability 1 � DN) and erroneously judge this item as ‘‘old’’ (with probability b) they may still guess that they‘‘remember’’ the item with probability R� or that they ‘‘know’’ the item with the complementary probability 1 � R�. In bothcases the parameters gDim1R

jA=B , gDim2RjA=B , gDim1K

jA=B , and gDim2KjA=B capture the guessing processes in source attributions.

All model-based analyses were conducted with data aggregated at the group level using the computer program multiTree(Moshagen, 2010). Based on previous studies (Meiser & Bröder, 2002) we expected item memory for words presented inlarge font to be superior to item memory for words presented in small font and therefore separate item recognition param-eters (D) were estimated for the two font size conditions. In order to render the model identifiable, distracter detection(parameter DN) was set equal to target recognition (parameter D) of the small font size condition as is common in two-highthreshold models of recognition memory (e.g., Meiser & Bröder, 2002).

The model provided a good fit to the data in the control group, G2(17) = 21.96, p = .19, and in the group with DT,G2(17) = 16.89, p = .46. Table 3 displays the parameter estimates.

Item recognition performance (D parameters) in the DT group was generally lower than in the control group,G2(2) = 84.43, p < .001. Furthermore, neither the probability of assigning R responses to recognized items (parameter R) dif-fered between the control group and the DT group, G2(1) = 1.77, p = .184, nor the probability of assigning R responses to erro-neously endorsed distracters and undetected old items (parameter R�), G2(1) = 0.56, p = .452.

Table 3Parameter values (and 95% confidence intervals) estimated with the multinomial model of multidimensional source memory in Experiments 1 and 2.

Parameter Experiment 1 Experiment 2

Control Dual task at encoding Control Dual task at encoding Dual task at retrieval

Dsmallfont .56 (.53; .59) .40 (.37; .43) .63 (.60; .65) .40 (.37; .43) .65 (.62; .67)Dlargefont .61 (.57; .64) .47 (.43; .51) .69 (.66; .71) .47 (.44; .51) .70 (.67; .73)R .76 (.73; .80) .73 (.69; .77) .74 (.72; .77) .68 (.65; .72) .74 (.72;77)R* .17 (.13; .22) .20 (.15; .24) .30 (.23; .38) .21 (.17; .26) .24 (.18; .30)b .31 (.28; .33) .28 (.26; .31) .18 (.15; .20) .29 (.26; .31) .27 (.24; .30)dR .12 (.05; .19) .15 (.05; .24) .20 (.13; .27) .14 (.06; .23) .15 (.09; .21)dK .00 (.00; .14)a .02 (.00; .16)a .06 (.00; .16)a .00 (.00; .15)a .10 (.00; .20)a

epositionR .09 (.00; .18) .09 (.00; .22)a .10 (.01; .20) .18 (.07; .29) .07 (.00; .15)a

epositionK .10 (.00; .28)a .00 (.00; .21)a .09 (.00; .22)a .19 (.02; .37) .04 (.00; .19)a

efontsizeR .17 (.09; .26) .30 (.20; .41) .30 (.23; .38) .19 (.08; .29) .25 (.18; .33)efontsizeK .16 (.00; .33)a .08 (.00; .29)a .24 (.11; .37) .08 (.00; .27)a .12 (.00; .26)a

Note. D = probability of recognizing a study word; R = probability of assigning a R response to a recognized studied word; R* = probability of assigning a Rresponse to distracter words that were guessed to be old; b = probability of classifying an undetected distracter as ‘‘old’’; d R = probability of joint contextmemory for study words classified as R; d K = probability of joint context memory for study words classified as K; e = independent memory for the contextfeatures, superscript indicates classification of memory experience.

a Confidence intervals with limits of .00 were bounded by the parameter space which limits the range of each model parameter to the interval (.00,1.00).

C.D. Boywitt, T. Meiser / Consciousness and Cognition 22 (2013) 875–886 881

In the next step, we compared the parameters dR and dK across groups to assess whether context–context binding wasaffected by DT load. There was no overall effect of DT on the binding parameters for R and K responses, G2(2) = 0.33,p = .847, but the binding parameters were generally larger for R than for K responses, G2(2) = 8.21, p = .017. Furthermore,the confidence intervals for the binding parameter for K responses in both groups overlapped with zero but they did notoverlap with zero for R responses, indicating that context binding was absent in K responses and present in R responsesin both groups.

In order to ensure that the lack of a difference in the binding parameter (dR) between the control group and the DT groupis not due to a lack of statistical power, we conducted a power analysis using the computer program multiTree (Moshagen,2010). The results indicate that the power for detecting an absolute difference of .20 (effect size w � .04) in the parameterestimates with the given sample (N = 9880 [104 items � 95 participants]) was 1 � b = .94. Thus, the statistical power wassufficient to detect substantial effects of the DT manipulation on context–context binding.

Last, comparing independent context memory for both dimensions and R and K responses (eDim1R, eDim1K, eDim2R, andeDim2K) between both groups revealed no significant overall differences in independent context memory, G2(4) = 6.92,p = .140.

4.3. Discussion

The results of the analysis of R and K responses are in line with previous studies which employed DT manipulations andelicited RK judgments: DT load selectively reduced the rate of R responses to target items and left the rate of K responsesunaffected (Gardiner, Gregg, Mashru, & Thaman, 2001; Gardiner & Parkin, 1990). Further, overall context memory (as mea-sured via ACSIM scores) was higher for R than for K responses in both groups, converging with results of other studies (e.g.,Dudukovic & Knowlton, 2006; Meiser & Bröder, 2002). Importantly, the similarity manipulation of the context features wassuccessful, as indicated by similar levels of context memory in the DT and the control group.

The model based analyses, however, suggest a different conclusion regarding the effect of DT load on RK-judgments.While the analysis of the raw frequencies of R and K responses indicates a lower rate of R responses in the DT group thanin the control group, the model-based analysis did not show any difference in the probability of assigning R responses torecognized items (parameter R). Note, however, that the raw frequencies of R and K responses are two-in-one measuresof item recognition and subjective retrieval experience because the observed rates of R and K responses add up to the totalrate of ‘‘old’’ responses. Accordingly, manipulations which affect item recognition performance or subjective retrieval expe-rience both manifest themselves in differences in R and K responses. The model-based analysis, on the other hand, decom-poses the observed frequency of R and K responses into the contributions of item recognition (parameter D), subjectiveretrieval experience (parameter R), and guessing processes (parameters b and R�). Thus, according to the model-based anal-ysis the observed difference in the rate of R responses was driven by differences in mere item recognition performance(parameter D) between the control group and the DT group and not by differences in genuine recollective experience(parameter R). Assuming that the recollective experience is associated with retrieval of some context details (Yonelinas,2002) this result can be interpreted as a consequence of the equating of context memory between the two groups: Becausethere were no differences in context memory between the DT and the control group, no differences in recollective experiencecould emerge.

Last but most importantly, the typical pattern of context–context binding for R but not for K responses was observed inthe control group and in the group with the DT load during encoding. This result suggests that context–context binding wasequally prone to reductions in available resources during encoding as was conscious recollection. Thus, DT load did not alterthe relation between context–context binding and conscious recollection.

5. Experiment 2

The central result of the first experiment is that context–context binding was not reduced by cognitive load during encod-ing over and above the effects on subjective retrieval experience. This leaves open the question whether a retrieval process asopposed to an encoding process might be involved in joint context memory (Meiser & Bröder, 2002; Starns & Hicks, 2008).For example, a cueing process between the context features (Meiser & Bröder, 2002) or between the item itself and each con-text feature (Starns & Hicks, 2008) during retrieval might produce stochastic dependency in context memory. The notion thatan additional (retrieval) process is involved in R responses is also compatible with extant dual process models of humanmemory (e.g., Jacoby, 1991; Mandler, 1980; Yonelinas, 1994). Most of these models assume that two different processesoperate at time of retrieval, a familiarity process on the one hand, and a recollection process on the other hand (see Yone-linas, 2002, for a review) and this latter process might also be involved in producing joint context memory. To test the role ofeffortful retrieval processes in producing context–context binding, we extended the design of Experiment 1 by including aDT group during retrieval. Second, we sought to look into the relation of resource-demanding processes in the subjectiveretrieval experience and context memory in general which could not be investigated in the first experiment because contextmemory was experimentally equated between groups. Although many previous studies have assessed the effects of dual taskload on RK judgments (Gardiner & Parkin, 1990; Gardiner et al., 2001; Yonelinas, 2001) or source memory (Craik et al., 2010;Troyer & Craik, 2000; Troyer et al., 1999) ours is the first study to our knowledge which jointly looks at the effects of dual

882 C.D. Boywitt, T. Meiser / Consciousness and Cognition 22 (2013) 875–886

task load on context memory for two orthogonal features and RK judgments. The parallel pattern of decreased rates of R re-sponses and decreased source memory performance in those previous studies suggests similar effects for both source mem-ory and recollective experience. What those studies do not reveal, however, is whether the relation of higher source memoryfor R than for K judgment is invariant across cognitive load conditions. Put more bluntly, is context memory still character-istic of R responses even under cognitive load? Indeed, there is evidence showing that this relationship may not be as uni-versal as suspected (Hicks, Marsh, & Ritschel, 2002; Meiser & Sattler, 2007).

5.1. Method

5.1.1. ParticipantsOne hundred and sixty-nine students (mean age = 20.97, SD = 2.16 years; 80% female) participated in this study for course

credit or monetary compensation and were randomly assigned to the three groups. Fifty-five participants were assigned tothe control group, 57 participants each were assigned to the dual task during encoding group and to the dual task duringretrieval group.

5.1.2. Materials and designThe materials were identical to the materials used in the first experiment with the following exceptions. The distance

between the left and right position on the screen was 19 cm between the centers of the stimuli in all groups. The large fontsize was 78 pt font and the smaller font size was 34 pt font in all groups.

The control group performed both the study and the test phase without a secondary task and one group received the addi-tional tone monitoring task during the study phase while another group received the tone monitoring task during the testphase.

5.1.3. ProcedureThe procedure was identical to the procedure in the first experiment with the following exceptions. Participants in the

group with the tone monitoring task during retrieval were informed about the additional task prior to the memory testand were given the same short practice block of the 8 tones as the group with the tone-monitoring task during encoding.The test block was divided into two parts of equal length with a break of one minute in between for all participants to avoidmotor fatigue effects in the group with the tone monitoring task during the test phase.

5.2. Results

5.2.1. Remember-know judgmentsTable 1 shows the proportions of R and K responses to target items (i.e., hits) and distracters (i.e., false alarms). Concern-

ing hits, a two by three mixed ANOVA with type of judgment (R vs. K) as within participants factor and DT group (control vs.load during encoding vs. load during retrieval) as between participants factor revealed a significant main effect of type ofjudgment, F(1,166) = 76.12, MSE = 0.051, g2

p = 0.31, p < .001, with higher hit rates for R than for K judgments. The main effectof DT group was also significant, F(2,166) = 13.70, MSE = 0.015, g2

p = 0.14, p < .001. However, the main effects were qualifiedby a significant interaction, F(2,166) = 8.98, MSE = 0.051, g2

p = 0.10, p < .001. Separate one-way ANOVAs on the hit rate with Rresponses and K responses revealed a main effect of DT group on the rate of R responses, F(2,166) = 12.51, MSE = 0.050,g2

p = 0.13, p < .001, but not on the rate of K responses, F(2,166) = 2.53, MSE = 0.016, g2p = 0.03, p = .083. Post hoc tests (LSD)

on the rate of R responses showed that the group with a secondary task during encoding produced significantly fewer R re-sponses than the control group, p < .001, and the group with the secondary task during retrieval, p < .001, but the controlgroup and the group with a secondary task during retrieval did not differ significantly, p = .72.

Concerning false alarms, a two by three mixed ANOVA with the factors judgment and DT group revealed a significantmain effect of type of judgment, F(1,166) = 67.36, MSE = 0.004, g2

p = 0.29, p < .001, with higher false alarm rates for K thanfor R responses, and a main effect of DT group, F(2,166) = 13.36, MSE = 0.006, g2

p = .14, p < .001. The interaction was also sig-nificant, F(2,166) = 8.95, MSE = 0.004, g2

p = 0.14, p < .001. Separate one-way ANOVAs revealed that the groups did not differ intheir false R rate, F(2,166) = 1.84, MSE = 0.003, g2

p = 0.02, p = .16, but differed in their false K rate, F(2,166) = 14.57,MSE = 0.008, g2

p = 0.15, p < .001. Post hoc tests (LSD) showed that this effect was due to the higher false K rate in the DT duringencoding group than in both other groups, p’s < .001, while false alarm rates in the control group did not differ from the falsealarm K rate in the DT during retrieval group p = .11.

These results suggest that the DT during encoding selectively reduced the rate of R responses to study items and left therate of K responses to study items unaffected. Similarly, false alarms increased only in the group with the DT during encodingselectively for K responses.

5.2.2. Overall context memoryAgain, overall context memory was above chance level for R responses in all groups (see Table 2).A mixed ANOVA on the ACSIM scores with the within participants factors context dimension (position vs. font size) and

judgment (R vs. K) and the between participants factor group (control vs. DT at encoding vs. DT at retrieval) revealed a maineffect of judgment, F(1,150) = 74.50, MSE = 0.02, g2

p = .33, p < .001, with higher context memory for R than for K responses.

C.D. Boywitt, T. Meiser / Consciousness and Cognition 22 (2013) 875–886 883

The main effect of context dimension was also significant, F(1,150) = 12.39, MSE = .02, g2p = .08, p = 001, with better context

memory for position than for font size. The main effect of group was marginally significant, F(2,150) = 2.98, MSE = 0.04,g2

p = .04, p = .054. Neither the interaction between judgment and DT group was significant, F(2,150) = 0.32, MSE = 0.02,g2

p = .004, p = .726, nor the three-way interaction, F(1,150) = 2.27, MSE = 0.02, g2p = .03, p = .107. The interaction between con-

text dimension and judgment, however, was significant, F(1,150) = 5.67, MSE = 0.02, g2p = .04, p = .019, as was the interaction

between context dimension and DT group, F(1,150) = 8.53, MSE = 0.02, g2p = .10, p < .001.

In order to probe these interactions, mixed ANOVAs with the within participants factor judgment (R vs. K) and the be-tween participants factor group (control vs. DT at encoding vs. DT at retrieval) were conducted separately for ACSIM scoresof font size and of position. For font size the analysis revealed a significant main effect of judgment, F(1,150) = 28.96,MSE = 0.02, g2

p = .16, p < .001, with higher context memory for R than for K responses but no main effect of DT group,F(2,150) = 0.02, MSE = 0.03, g2

p = .008, p = .559, and no interaction, F(1,150) = 1.04, MSE =.02, g2p = .01, p = .356. The same AN-

OVA on ACSIM scores for position, however, revealed in addition to the main effect of judgment, F(1,150) = 56.56, MSE = 0.02,g2

p = .27, p < .001, with better context memory for R than for K judgments, a significant effect of DT group, F(2,150) = 8.22,MSE = 0.03, g2

p = .10, p < .001, but no interaction, F(1,150) = 1.25, MSE = 0.02, g2p = .02, p =.29. Post hoc tests (LSD) showed that

context memory in the DT at encoding group was lower than in the control group, p < .001, and lower than in the DT at re-trieval group, p = .019, but the latter two groups did not differ significantly, p =.111.

5.2.3. Model-based analysesThe multinomial model in Fig. 1 provided an adequate fit to the data in the control group, G2(17) = 26.06, p = .07, and a

good fit in the group with the DT during encoding, G2(17) = 6.39, p = .99, and the group with the DT during retrieval,G2(17) = 19.84, p = .28. Refer to Table 3 for the parameter estimates.

The main effect of DT group on item recognition (D parameters) was significant, G2(4) = 348.77, p < .001. This effect wasdriven by overall lower recognition performance in the DT during encoding group than in the control group, G2(2) = 235.25,p < .001, and in the dual task during retrieval group, G2(2) = 285.08, p < .001, while the latter two groups did not differ sig-nificantly, G2(2) = 1.98, p = .372.

The overall test of the equivalence of the probability of assigning R responses to recognized target items (R parameter)was also significant, G2(2) = 7.94, p = .019. Post hoc comparisons revealed no significant difference in the R parameter be-tween the DT at retrieval group and the control group, G2(1) = 0.01, p = .94, but lower estimates in the DT at encoding groupthan in the control group, G2(1) = 7.19, p = .007, and the DT at retrieval group, G2(1) = 6.74, p = .009.

The probability of assigning an R response to erroneously endorsed distracters or undetected study items (parameter R�)did not differ significantly between the three groups, G2(2) = 4.34, p = .114.

Regarding, context–context binding, there were neither overall differences in the binding parameter for R responses (dR)between the three groups, G2(2) = 1.65, p = .438, nor in the binding parameter for K responses (dK), G2(2) = 1.74, p = .419.However, the binding parameters for R responses were generally larger than the binding parameters for K responses,G2(3) = 9.96, p =.019. Furthermore, the confidence intervals for the binding parameter for K responses but not for R responsesoverlapped with zero in all groups. Thus, context–context binding was again absent in K responses and present in R re-sponses for all groups.

Comparing independent context memory for R and K responses for both context dimensions (parameters eDim1R, eDim1K,eDim2R, and eDim2K) across groups revealed significant differences, G2(8) = 19.27, p = .013. Post hoc tests showed that indepen-dent context memory differed significantly between the DT during encoding group and the control group, G2(4) = 16.62,p = .002, and the DT during retrieval group, G2(4) = 12.01, p = .017. But the latter two groups did not differ in independentcontext memory, G2(4) = 2.39, p = .665. Probing this main effect, we compared context memory for each context dimensionand R and K responses separately between the dual task condition during encoding and both other conditions jointly. Due todecreased degrees of freedoms for these analyses none of the post hoc comparisons was statistically significant, epositionK:G2(1) = 2.03, p = .154, epositionR: G2(1) = 2.18, p = .139, efontsizeK, G2(1) = 2.11, p = .146, efontsizeR, G2(1) = 2.19, p = .139.

5.3. Discussion

To summarize, the DT load during encoding yielded a lower rate of R responses but left K responses unaffected while DTduring retrieval had almost no effect on RK judgments. Again, this pattern of results is well in line with previous studies (e.g.,Gardiner & Parkin, 1990; Gardiner et al., 2001; Yonelinas, 2001) and highlights the importance of encoding processes for thesubsequent recollective experience. A novel finding in this regard is that the DT load during retrieval did not have any det-rimental effects on either the rates of R or K responses.

The analyses of overall context memory (as measured via ACSIM scores) revealed a selective reduction of context memoryin the DT during encoding group but not in the DT during retrieval group, in line with previous studies (e.g., Craik et al., 2010;Troyer et al., 1999). Importantly, the relation between context memory and subjective retrieval experience such that R re-sponses are associated with higher context memory than K responses was preserved even under DT load. Furthermore,the selective context memory impairment due to dual task load during encoding was limited to the dimension position. Thisresult is consistent with the idea that context dimensions may differ with regard to the processing demands such that stim-ulus-bound (i.e., intrinsic) context features are less prone to DT manipulations than spatio-temporal (i.e., extrinsic) contextfeatures (Moscovitch, 1992; Spencer & Raz, 1995; Troyer et al., 1999). According to this reasoning, position on the screen can

884 C.D. Boywitt, T. Meiser / Consciousness and Cognition 22 (2013) 875–886

be considered a spatial–temporal feature which requires more effort to encode than font size which can be considered amore stimulus-bound feature. Reducing the available resources during encoding thus affected the resource-demandingencoding of position but not the encoding of font size.

In line with the results of the analysis of the behavioral data, the model-based analyses revealed that item recognition (Dparameter) as well as recollective experience (R parameter) were significantly reduced by the DT load during encoding butnot by DT load during retrieval. Unlike in Experiment 1, the difference in the observed rate of R responses between the DTload during encoding group and the control group appeared to have been driven not only by differences in item recognitionperformance (D parameter) but also by differences in genuine recollective experience (R parameter). In combination withlower overall context memory in the DT load during encoding group this result suggests that decreasing context memorydecreases recollective experience. Crucially, context–context binding was still present for R but not for K responses in allthree groups and did not differ between the DT groups and the control group even though overall context memory was high-er, at least for the dimension position, in the control group than in the DT during encoding group.

These results corroborate the conclusions drawn from the first experiment that context–context binding of perceptualfeatures does not rely on resource-demanding processing over and above the demands of conscious recollection. Thus, giventhe experience of consciously recollected memories, the quality of associated context memory did not differ across the threegroups although this experience occurred substantially less frequently if a cognitive load was imposed during encoding.

6. General discussion

A substantial body of research indicates that the subjective experience of conscious recollection is associated with sub-stantial memory for the context details of the encoding episode (e.g., Dewhurst & Hitch, 1999; Dudukovic & Knowlton, 2006;Meiser & Bröder, 2002; Perfect et al., 1996) and, more importantly, with bound memory of these context features (Meiser &Bröder, 2002; Meiser & Sattler, 2007; Meiser et al., 2008). Recently, it has been suggested that context–context binding isaccomplished by an encoding process operating in the focus of attentional processing which integrates the stimulus andits context features into a coherent memory representation of the feature configuration during study (Boywitt & Meiser,2012b; Uncapher et al., 2006). In line with this suggestion, research on short-term memory feature binding offers ample evi-dence that the integration of multiple perceptual features is resource-demanding (e.g., Stefurak & Boynton, 1986; Treisman &Schmidt, 1982; Wheeler & Treisman, 2002). Contrary to this hypothesis, however, in the first experiment context–contextbinding was present in R responses even under DT load during encoding when overall context memory was equated betweenthe DT group and the control group. The second experiment corroborated this finding and extended it to DT manipulationsduring retrieval. In both experiments context–context binding was correlated with the recollective experience which wasonly affected by DT load during encoding but not by DT load during retrieval. Thus, context–context binding appears tobe associated with the processes driving conscious recollection but there was no evidence that effortful processing overand above the processing for recollection is involved in context binding. This result is well in line with previous studies sug-gesting that the processes underlying context memory and R responses may be largely identical. For example, the rate of Rresponses rises with the distinctiveness of sources, which in turn enhances memory for context (Conway & Dewhurst, 1995;Donaldson, MacKenzie, & Underhill, 1996). Furthermore, activity in the same brain areas, such as the hippocampus and pre-frontal regions, appears to be characteristic of both context memory and R responses, (e.g., Eldridge, Knowlton, Furmanski,Bookheimer, & Engel, 2000; Yonelinas, Otten, Shaw, & Rugg, 2005). Contrary evidence, however, can be found in studies inwhich R responses and context memory were dissociated (e.g., Hicks et al., 2002; Meiser & Sattler, 2007). Thus the conclusionthat both context memory and recollection are the product of the same process may be premature.

Furthermore, other DT manipulations might be more successful in dissociating context–context binding and the recollec-tive experience. For instance, secondary tasks which tap visuo-spatial processing as opposed to purely central processingmight be more effective in impairing binding of visuo-spatial features. On the other hand, the evidence for resource-demand-ing processing in feature binding in visual short-term memory is not unequivocal (see Baddeley, Allen, & Hitch, 2011, for areview). Thus there might be moderators for the role of effortful processing in feature binding. One likely candidate iswhether the features are stimulus-bound or spatio–temporal. Evidence from differences in the processing demands of stim-ulus-bound and spatio-temporal features comes from earlier studies (Spencer & Raz, 1995; Troyer et al., 1999) but can alsobe found here. In the second experiment, in which context memory was allowed to vary between groups, only memory forposition on the screen but not memory for font size was affected by the DT load. Interestingly, concepts similar to the dis-tinction between stimulus-bound and spatial temporal features have been proposed in the visual cognition literature underthe rubric of unitization effects (Ecker, Maybery, & Zimmer, 2013; Walker & Cuthbert, 1998). The context features used herecould be seen as a combination of stimulus-bound (i.e., font size) and spatio-temporal (i.e., position on the screen) and ittherefore remains an interesting question for future research whether binding of combinations of purely spatio-temporalcontext features are more prone to dual-task manipulations. In line with the distinction between stimulus-bound and spa-tio-temporal context features the present results can be taken as an indication that different context features may also differin their role for the subjective retrieval experience. While memory for the rather stimulus-bound feature font size was notprone to the DT manipulation, memory for the spatial temporal feature position was reduced by DT load and this effect wasparallel to the effect of DT load on recollective experience. Future research should thus investigate whether differences insubjective retrieval experience could be related to differences in the processing of different context details.

C.D. Boywitt, T. Meiser / Consciousness and Cognition 22 (2013) 875–886 885

The evidence for the binding of context features in episodic memory, however, goes beyond the demonstration of visualfeature binding. Previous research has shown that conscious recollection is associated with cross-modal binding. Meiser andcolleagues (Meiser & Sattler, 2007; Meiser et al., 2008), for instance, used visual context features such as font size and anauditory semantic feature such as gender of a speaker to show that binding of these features too was characteristic of con-scious recollection. Whereas it is possible that visuo-spatial features might be bound in rather early stages of processing (seeLuck & Vogel, 1997), features of different modalities might be more likely to rely on a central binding process. This idea re-ceives preliminary support by a study showing that prefrontal activity did not differ between a condition in which two visualfeatures were part of the same object and a condition in which two visual features were spatially separated (Kuo & Van Pet-ten, 2008). Parietal activity, however, was correlated with context memory performance, indicating that visual processingwas mainly involved in binding of these visual features. Future research should thus extend the investigation to other com-binations of context features.

7. Conclusions

Regarding the role of overall context memory for the recollective experience, the present research offers original evidencefor the causal role of context memory in producing the subjective experience of conscious recollection by looking at DT ef-fects on the subjective retrieval experience and context memory jointly. In Experiment 1 experimentally equated similar lev-els of overall context memory in the DT group and the control group were accompanied by similar levels of recollectiveexperience as indicated by model-based analyses. In Experiment 2, on the other hand, differences in overall context memorybetween the DT during encoding group and the control group were accompanied by significant differences in the recollectiveexperience such that lower levels of context memory went together with lower levels of recollective experience. Impor-tantly, however, the general relation of higher context memory for R than for K responses, however, was preserved acrossboth experiments as well as under cognitive load during encoding or retrieval.

While cognitive load during encoding leads to reductions in the frequency of the subjective experience of conscious rec-ollection and reductions in overall context memory, context–context binding was not affected over and above the effects onconscious recollection. Thus, given participants report recollective retrieval experiences, there was no difference in the qual-ity of the associated memory characteristics in terms of context memory and context–context binding.

References

Allen, R. J., Baddeley, A. D., & Hitch, G. J. (2006). Is the binding of visual features in working memory resource-demanding? Journal of ExperimentalPsychology: General, 135, 298–313.

Baddeley, A. D., Allen, R. J., & Hitch, G. J. (2011). Binding in visual working memory: The role of the episodic buffer. Neuropsychologia, 49, 1393–1400.Batchelder, W. H., & Riefer, D. M. (1990). Multinomial processing models of source monitoring. Psychological Review, 97, 548–564.Bayen, U. J., Murnane, K., & Erdfelder, E. (1996). Source discrimination, item detection, and multinomial models of source monitoring. Journal of Experimental

Psychology: Learning, Memory, and Cognition, 22, 197–215.Boywitt, C. D., Kuhlmann, B. G., & Meiser, T. (2012). The role of source memory in older adults’ recollective experience. Psychology and Aging, 27, 484–497.Boywitt, C. D., & Meiser, T. (2012a). Bound context features are integrated at encoding. The Quarterly Journal of Experimental Psychology, 1484–1501.Boywitt, C. D., & Meiser, T. (2012b). The role of attention for context–context binding of intrinsic and extrinsic features. Journal of Experimental Psychology:

Learning, Memory, and Cognition, 38, 1099–1107.Brown, L. A., & Brockmole, J. R. (2010). The role of attention in binding visual features in working memory: Evidence from cognitive ageing. The Quarterly

Journal of Experimental Psychology, 63, 2067–2079.Conway, M. A., & Dewhurst, S. A. (1995). Remembering, familiarity, and source monitoring. The Quarterly Journal of Experimental Psychology, 48, 125–140.Craik, F. I. M., Luo, L., & Sakuta, Y. (2010). Effects of aging and divided attention on memory for items and their contexts. Psychology and Aging, 25, 968–979.Delvenne, J. F., Cleeremans, A., & Laloyaux, C. (2010). Feature bindings are maintained in visual short-term memory without sustained focused attention.

Experimental Psychology, 57, 108–116.Dewhurst, S. A., & Hitch, G. J. (1999). Cognitive effort and recollective experience in recognition memory. Memory, 7, 129–146.Donaldson, W., MacKenzie, T. M., & Underhill, C. F. (1996). A comparison of recollective memory and source monitoring. Psychonomic Bulletin & Review, 3,

486–490.Dudukovic, N. M., & Knowlton, B. J. (2006). Remember-know judgments and retrieval of contextual details. Acta Psychologica, 122, 160–173.Ecker, U. K. H., Maybery, M., & Zimmer, H. D. (2013). Binding of intrinsic and extrinsic features in working memory. Journal of Experimental Psychology:

General, 142, 218–234.Ecker, U. K. H., Zimmer, H. D., & Groh-Bordin, C. (2007). Color and context: An ERP study on intrinsic and extrinsic feature binding in episodic memory.

Memory & Cognition, 35, 1483–1501.Eldridge, L. L., Knowlton, B. J., Furmanski, C. S., Bookheimer, S. Y., & Engel, S. A. (2000). Remembering episodes: A selective role for the hippocampus during

retrieval. Nature Neuroscience, 3, 1149–1152.Elsley, J. V., & Parmentier, F. B. R. (2009). Is verbal–spatial binding in working memory impaired by a concurrent memory load? The Quarterly Journal of

Experimental Psychology, 62, 1696–1705.Erdfelder, E., Auer, T. S., Hilbig, B. E., Assfalg, A., Moshagen, M., & Nadarevic, L. (2009). Multinomial processing tree models a review of the literature.

Zeitschrift für Psychologie/Journal of Psychology, 217, 108–124.Gardiner, J. M., Gregg, V. H., Mashru, R., & Thaman, M. (2001). Impact of encoding depth on awareness of perceptual effects in recognition memory. Memory

& Cognition, 29, 433–440.Gardiner, J. M., & Parkin, A. J. (1990). Attention and recollective experience in recognition memory. Memory & Cognition, 18, 579–583.Hicks, J. L., Marsh, R. L., & Ritschel, L. (2002). The role of recollection and partial information in source monitoring. Journal of Experimental Psychology:

Learning, Memory, and Cognition, 28, 503–508.Jacoby, L. L. (1991). A process dissociation framework – Separating automatic from intentional uses of memory. Journal of Memory and Language, 30,

513–541.Kuo, T. Y., & Van Petten, C. (2008). Perceptual difficulty in source memory encoding and retrieval: Prefrontal versus parietal electrical brain activity.

Neuropsychologia, 46, 2243–2257.

886 C.D. Boywitt, T. Meiser / Consciousness and Cognition 22 (2013) 875–886

Light, L. L., & Berger, D. E. (1976). Are there long-term literal copies of visually presented words. Journal of Experimental Psychology: Human Learning andMemory, 2, 654–662.

Luck, S. J., & Vogel, E. K. (1997). The capacity of visual working memory for features and conjunctions. Nature, 390, 279–281.Mandler, G. (1980). Recognizing – The judgment of previous occurrence. Psychological Review, 87, 252–271.Meiser, T. (2009, October). Multinomial analysis of multidimensional source memory: Overview and new extensions. Paper presented at the Winer Lectures,

Purdue University, West Lafayette.Meiser, T., & Bröder, A. (2002). Memory for multidimensional source information. Journal of Experimental Psychology: Learning, Memory, and Cognition, 28,

116–137.Meiser, T., & Sattler, C. (2007). Boundaries of the relation between conscious recollection and source memory for perceptual details. Consciousness and

Cognition, 16, 189–210.Meiser, T., Sattler, C., & Weisser, K. (2008). Binding of multidimensional context information as a distinctive characteristic of remember judgments. Journal

of Experimental Psychology: Learning, Memory, and Cognition, 34, 32–49.Moscovitch, M. (1992). Memory and working-with-memory – A component process model based on modules and central systems. Journal of Cognitive

Neuroscience, 4, 257–267.Moshagen, M. (2010). Multitree: A computer program for the analysis of multinomial processing tree models. Behavior Research Methods, 42, 42–54.Mulligan, N. W. (2011). Generation disrupts memory for intrinsic context but not extrinsic context. The Quarterly Journal of Experimental Psychology, 64,

1543–1562.Murnane, K., & Bayen, U. J. (1996). An evaluation of empirical measures of source identification. Memory & Cognition, 24, 417–428.Parkin, A. J., & Russo, R. (1990). Implicit and explicit memory and the automatic effortful distinction. European Journal of Cognitive Psychology, 2, 71–80.Parkin, A. J., & Walter, B. M. (1992). Recollective experience, normal aging, and frontal dysfunction. Psychology and Aging, 7, 290–298.Perfect, T. J., Mayes, A. R., Downes, J. J., & VanEijk, R. (1996). Does context discriminate recollection from familiarity in recognition memory? The Quarterly

Journal of Experimental Psychology, 49, 797–813.Rajaram, S. (1993). Remembering and knowing – Two means of access to the personal past. Memory & Cognition, 21, 89–102.Rensink, R. A. (2000). The dynamic representation of scenes. Visual Cognition, 7, 17–42.Riefer, D. M., Chien, Y., & Reimer, J. F. (2007). Positive and negative generation effects in source monitoring. The Quarterly Journal of Experimental Psychology,

60, 1389–1405.Spencer, W. D., & Raz, N. (1995). Differential effects of aging on memory for content and context: A meta-analysis. Psychology and Aging, 10, 527–539.Starns, J. J., & Hicks, J. L. (2008). Context attributes in memory are bound to item information, but not to one another. Psychonomic Bulletin & Review, 15,

309–314.Stefurak, D. L., & Boynton, R. M. (1986). Independence of memory for categorically different colors and shapes. Perception & Psychophysics, 39, 164–174.Treisman, A., & Gelade, G. (1980). Feature-integration theory of attention. Cognitive Psychology, 12, 97–136.Treisman, A., & Schmidt, H. (1982). Illusory conjunctions in the perception of objects. Cognitive Psychology, 14, 107–141.Troyer, A. K., & Craik, F. I. M. (2000). The effect of divided attention on memory for items and their context. Canadian Journal of Experimental Psychology, 54,

161–171.Troyer, A. K., Winocur, G., Craik, F. I. M., & Moscovitch, M. (1999). Source memory and divided attention: Reciprocal costs to primary and secondary tasks.

Neuropsychology, 13, 467–474.Tulving, E. (1985). Memory and consciousness. Canadian Psychology, 26, 1–12.Uncapher, M. R., Otten, L. J., & Rugg, M. D. (2006). Episodic encoding is more than the sum of its parts: An fmri investigation of multifeatural contextual

encoding. Neuron, 52, 547–556.Walker, P., & Cuthbert, L. (1998). Remembering visual feature conjunctions: Visual memory for shape-colour associations is object-based. Visual Cognition, 5,

409–455.Wheeler, M. E., & Treisman, A. M. (2002). Binding in short-term visual memory. Journal of Experimental Psychology: General, 131, 48–64.Yonelinas, A. P. (1994). Receiver-operating characteristics in recognition memory – Evidence for a dual-process model. Journal of Experimental Psychology:

Learning, Memory, and Cognition, 20, 1341–1354.Yonelinas, A. P. (2001). Consciousness, control, and confidence: The 3 Cs of recognition memory. Journal of Experimental Psychology-General, 130, 361–379.Yonelinas, A. P. (2002). The nature of recollection and familiarity: A review of 30 years of research. Journal of Memory and Language, 46, 441–517.Yonelinas, A. P., Otten, L. J., Shaw, K. N., & Rugg, M. D. (2005). Separating the brain regions involved in recollection and familiarity in recognition memory.

Journal of Neuroscience, 25, 3002–3008.